Robinhood hands AI agents the keys to its trading platform and credit card Proactive uses images sourced from Shutterstock
Robinhood Markets Inc (NASDAQ:HOOD), the US retail brokerage, has launched a service that allows artificial intelligence agents to autonomously trade stocks and make credit card purchases on behalf of its 27.5 million customers.
The move, announced on Wednesday, represents one of the first attempts by a major financial platform to bring so-called agentic AI, where software acts independently rather than simply responding to prompts, directly into retail investing and consumer spending.
Robinhood shares rose about 3% on the news and continued to gain in after-hours trading.
The service, called Agentic Trading, works through a dedicated account that sits entirely separate from a customer’s main portfolio.
Users deposit a specific amount of capital into this sandboxed account and then connect a third-party AI agent, built on platforms such as Anthropic’s Claude or OpenAI’s ChatGPT, via Robinhood’s Model Context Protocol (MCP) servers.
The agent can then place trades, monitor positions and execute strategies without the user initiating each transaction.
Robinhood outlined several use cases: a long-term investor might instruct an agent to rebalance their portfolio when sector concentrations drift beyond set thresholds, while a more active trader could deploy a mean-reversion strategy with backtesting built in.
The feature launches in beta and is limited to equities for now. Robinhood said it plans to expand into options, cryptocurrency, event contracts, futures and prediction markets following the testing phase.
Alongside the trading product, Robinhood also introduced an agentic credit card.
Gold Card customers can link an AI agent to a dedicated virtual card with its own spending limit, allowing the agent to monitor prices, track product availability and make purchases autonomously. Users can choose whether to require manual approval for each transaction or let the agent spend freely within the cap.
The company was direct about the risks involved. Users bear full responsibility for any outcomes, and Robinhood does not supervise, control or guarantee the performance of any connected agent. The firm acknowledged that AI agents can misinterpret instructions, act on incomplete or stale data and behave unpredictably, potentially losing the full amount deposited.
The regulatory dimension is significant. FINRA’s 2026 regulatory oversight report flagged autonomous AI agents as requiring novel supervision frameworks, including tracking agent actions and restricting system access.
That guidance arrived just months before Robinhood’s launch, and whether the existing regulatory architecture around suitability obligations and best execution can accommodate software making unsupervised investment decisions remains an open question.